Checkbox Ethics ✅️

What does it look like in real terms?

Checkbox ethics is one of the most important phenomena to understand if you want to do the work of AI Governance properly, because the gap between policy and practice is where real harm gets done.

So what does it actually look like, in real terms?

Picture an organisation that publishes an AI ethics policy, complete with all the right principles around fairness, transparency and accountability.

The document is well written. It probably took someone weeks to draft. And then it gets parked on the intranet. Nobody in procurement has read it when they sign the next vendor contract. The development team doesn't know it exists. The board approved it months ago and considers the matter closed. The document is real. The governance is not.

Or take bias auditing. A third party gets commissioned to run an audit at the point of initial deployment, and their report concludes the system performs acceptably. Two years pass. The model has been updated. The deployment context has shifted. And the communities most affected by its outputs have never been asked a single question about how it's actually landing in their lives. The audit happened. The accountability did not.

You see the same pattern in psychological safety work. A leadership team sits through a half-day workshop, generates a list of team commitments, and someone prints them on a poster for the break room. Six months later, the most senior person in the room still visibly shuts down challenge the moment it appears. Junior staff still don't raise concerns before they become crises. And the poster is still there, on the wall, doing nothing. The training happened. The culture did not change.

Maybe think about that same manager carrying out an annual performance review. They tick the wellbeing box and consider pastoral care handled for the year. The team member who raised something genuinely difficult in that conversation gets no follow-up, no structural change, nothing. What they learn instead is that raising things produces paperwork, not action.

Check box ethics also shows up in inclusion and representation too. A governance body adds one person from an underrepresented community to a working group. That person is outvoted on every substantive question that matters. Their dissent gets noted in the minutes, and then nothing happens to it. The organisation can now report that diverse voices were consulted. The decision underneath remains exactly what it would have been anyway.

What ties every one of these together is this. Checkbox ethics confuses the production of an artefact with the achievement of a purpose. The policy, the training, the audit, the consultation, the report, all of it gets treated as the outcome, when it was only ever meant to be an instrument toward one.

The question checkbox ethics never asks is whether anything actually changed for the people that artefact was supposed to protect.

That is the question I put back at the centre of my work. So if you'd like a conversation about what that looks like inside your organisation, get in touch. DM me, or email louise@kairosynthesis.com.